Forecasting monthly streamflow to long-lead times: a preliminary study

[摘要] The availability of long range streamflow forecasts is highly valuable for water resources management. The Australian Bureau of Meteorology currently provides monthly updates of probabilistic forecasts of total streamflow volumes for the following three months at over 70 sites across Australia. In a recent survey conducted by the Bureau, key water agencies were very positive about the current seasonal streamflow forecasting service. However, many agencies have expressed strong interest in having the forecasts extended to longer lead times and having the three month total volumes broken down into monthly volumes. Beyond skilful forecast lead times, the water agencies are also very interested in stochastic scenarios out to even longer lead times, which can be used in conjunction with the skilful forecasts as inputs to water allocation models for forward planning and operations.In this paper, we report our development of a method for generating forecast guided stochastic scenarios. The method first generates rainfall forecasts statistically using seasonal climate model predictions of rainfall and sea surface temperature as predictors. The rainfall forecasts are then used as inputs to a monthly water balance model, followed by error updating and uncertainty quantification, to produce ensemble forecasts of monthly streamflow time series. The method extends the current forecast horizon of three months to when skilful forecasts can be obtained and transit the forecasts to stochastic scenarios as skill diminishes at longer lead times. Test results on three alpine catchments in south east Australia show that the forecasts are skilful to lead times of at least 2 months, and up to 5 months, depending on time of year. At longer lead times, skill scores are close to zero as forecasts revert to historically observed frequency distributions. The forecasts are shown to be statistically reliable in representing forecast uncertainty.

[发布日期] 2013-06-30 [发布机构] CSIRO

[学科分类] 地球科学(综合)